Using Image Filters in Pillow

filter() returns a new image. It does not normally replace the pixels in img. Assigning the result to a separate variable preserves the original image and makes it possible to compare both versions.

show() previews an image using an available image viewer. For a durable result, use save() and provide a different output filename.

Common Built-in Pillow Filters

The visible result depends on the source image. A detailed photograph may show filter differences clearly, while a plain image with little texture or contrast may not.

Example: Apply a Blur Filter

BLUR spreads local image information, reducing fine detail and producing a soft-focus effect.

from PIL import Image, ImageFilter

img = Image.open('input.jpg')
blurred = img.filter(ImageFilter.BLUR)

blurred.show()
blurred.save('input_blurred.png')

The original img remains available. The blurred variable refers to the new processed image.

Example: Detect Image Edges

FIND_EDGES highlights places where pixel intensity changes. These changes often correspond to object boundaries, outlines, or strong transitions between light and dark areas.

from PIL import Image, ImageFilter

img = Image.open('input.jpg')
image_edges = img.filter(ImageFilter.FIND_EDGES)

image_edges.show()
image_edges.save('input_edges.png')

The result is useful for inspecting outlines, but it is not the same as a complete object-recognition system. It identifies visual transitions rather than understanding what an object is.

Convert a Color Image to Grayscale

Grayscale conversion is different from filtering. Use the Image.convert() method, not ImageFilter, to change the image color mode.

from PIL import Image

img = Image.open('input.jpg')
black_white_image = img.convert('L')

black_white_image.show()
black_white_image.save('input_grayscale.png')

The mode 'L' means luminance. An 'L' image stores brightness values rather than separate red, green, and blue channels. It can contain many shades between dark and light, so it is grayscale but not necessarily a two-color black-and-white image.

If an application requires only pure black and pure white pixels, grayscale conversion alone is not enough. A separate thresholding or binary-conversion workflow is required.

Compare Several Filters

Applying multiple filters independently to one source image makes their differences easier to recognize.

from PIL import Image, ImageFilter

img = Image.open('input.jpg')

blurred = img.filter(ImageFilter.BLUR)
sharpened = img.filter(ImageFilter.SHARPEN)
embossed = img.filter(ImageFilter.EMBOSS)
edges = img.filter(ImageFilter.FIND_EDGES)

blurred.save('result_blur.png')
sharpened.save('result_sharpen.png')
embossed.save('result_emboss.png')
edges.save('result_edges.png')

• Compare BLUR and SHARPEN to see how softness and edge definition differ.
• Use EMBOSS to inspect how shading can suggest raised surfaces.
• Use FIND_EDGES to focus on boundaries rather than normal color and texture.

Chain Image Operations

Because filtering returns an image, the returned image can be used as the input to another operation. This is called chaining.

from PIL import Image, ImageFilter

img = Image.open('input.jpg')

processed = img.filter(ImageFilter.SHARPEN)
processed = processed.convert('L')
processed = processed.filter(ImageFilter.FIND_EDGES)

processed.show()
processed.save('sharpened_grayscale_edges.png')

Each statement creates or returns an image for the next step. Preserve the original by continuing to use separate variables or by never assigning processed results back to the source variable.

A Typical Filter Workflow

1. Import Image and, when needed, ImageFilter.
2. Open the source file with Image.open().
3. Apply a filter with filter(), or convert the color mode with convert('L').
4. Inspect the result with show() or another image viewer.
5. Save the processed image under a different filename.

from PIL import Image, ImageFilter

source = Image.open('input.jpg')
result = source.filter(ImageFilter.DETAIL)
result.save('input_detail.png')

Troubleshooting

ImageFilter is not defined

Import the module before referring to its filters:

from PIL import ImageFilter

Without this import, expressions such as ImageFilter.BLUR and ImageFilter.FIND_EDGES cannot be resolved.

The original image appears unchanged

The result of filter() may have been discarded. Store it and then display or save that variable:

processed_image = img.filter(ImageFilter.BLUR)
processed_image.show()

The source image was overwritten

Saving to the same path as the input can replace the original file. Use a distinct output path such as input_blurred.png while testing.

The grayscale image is not only black and white

convert('L') creates luminance grayscale with many possible brightness levels. It does not create a strictly two-color image. Use a separate thresholding or binary conversion process when only black and white pixels are wanted.

Filter effects are hard to see

Try a photograph containing textures, objects, and strong boundaries. Also save several results from the same source and compare them side by side. Effects vary with image content, contrast, size, and detail.

Key Terms

• Pillow: A Python imaging library for manipulating image files.
• PIL: The import namespace used by Pillow.
• ImageFilter: The Pillow module containing predefined image filters.
• Image.filter(): An image method that applies a filter and returns a processed image.
• Image.convert(): An image method that converts an image to a specified color mode.
• RGB: A color model using red, green, and blue channels.
• Grayscale: An image representation based on brightness rather than full color.
• 'L' mode: Pillow's luminance mode for grayscale images.

Exam-Ready Summary

• Import built-in filters with from PIL import ImageFilter.
• Open an image first; filters operate on an existing Pillow image object.
• Apply a filter with img.filter(filter_object).
• filter() returns a new image, so assign its result.
• Use show() to preview and save() to write output.
• Use img.convert('L') for luminance grayscale, not ImageFilter.
• Use separate filenames to preserve the original image.

For a compact reference, see Using Image Filters.